Bug 1900656

Created attachment 1732541 [details]
File: backtrace

Created attachment 1732542 [details]
File: core_backtrace

Created attachment 1732543 [details]
File: cpuinfo

Created attachment 1732544 [details]
File: dso_list

Created attachment 1732545 [details]
File: environ

Created attachment 1732546 [details]
File: limits

Created attachment 1732547 [details]
File: maps

Created attachment 1732548 [details]
File: mountinfo

Created attachment 1732549 [details]
File: open_fds

Created attachment 1732550 [details]
File: proc_pid_status

Created attachment 1732551 [details]
File: var_log_messages

Looks like we crash here when memory allocation fails. (I think that is only possible if you have modified kernel settings. Have you?)

I think a crash is inevitable if allocations are failing, because g_malloc() will crash if allocation fails, but this particular crash is just a bug that we can fix. It should crash later instead.

Thanks for reporting. Proposed: https://gitlab.gnome.org/GNOME/epiphany/-/merge_requests/863

Nope. Nothing modified. Standard kernel and no special command line:

> $ cat /proc/cmdline 
> BOOT_IMAGE=(hd1,gpt2)/vmlinuz-5.9.8-100.fc32.x86_64 root=/dev/mapper/fedora_ossman-root ro rd.lvm.lv=fedora_ossman/root rhgb quiet

This started happening after an upgrade from Fedora 31 to 32. All the web applications work fine up until the crash. They also always crash in the background, never when I'm actively using them.

Strange. Are you actually running out of memory?

Nope. Not as far as I can tell. Currently sitting at 9 GiB available and it's been crashing on and off constantly.

It seems to get more unstable the more web apps I have open. One seems fine, two crashes now and then and three crashes fairly quickly.

Yet comment #11 suggests it is out of memory. Or at least *thinks* it is.

Anyway, my merge request should downgrade this into a soft failure of the safe browsing feature: phishing checks will just be disabled when this happens, instead of crashing. (That's going to be the case in the future anyway, due to changes in Google's terms of service for their Safe Browsing API.)

Fair enough. What's the time frame for this fix to trickle down to Fedora 32? It's crashing a lot for me right now so some fix would be nice.

And in case it helps, here are the lines before it started complaining about memory:

> Nov 23 13:51:50 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute delete full hash statement: database is locked
> Nov 23 13:52:36 ossman.lkpg.cendio.se epiphany[3509]: Failed to execute clear hash prefix statement: attempt to write a readonly database
> Nov 23 13:52:48 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute update threat statement: attempt to write a readonly database
> Nov 23 13:52:59 ossman.lkpg.cendio.se epiphany[3509]: Failed to commit transaction on GSB database: disk I/O error
> Nov 23 13:53:12 ossman.lkpg.cendio.se epiphany[3509]: Failed to execute update threat statement: attempt to write a readonly database
> Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: attempt to write a readonly database
> Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory
> Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory
> Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory

Those are the first lines from epiphany since upgrading to Fedora 32.

(In reply to Pierre Ossman from comment #18)
> Fair enough. What's the time frame for this fix to trickle down to Fedora
> 32? It's crashing a lot for me right now so some fix would be nice.

Hm, I don't think it's important enough for a F32 update, sorry. I'll backport it to 3.36 since you requested it, so it will be included if there are any further updates. For now, you can avoid this crash by going into privacy settings and disabling Block Dangerous Websites.

It should reach F33 pretty quickly, though.

> And in case it helps, here are the lines before it started complaining about
> memory:
> 
> > Nov 23 13:51:50 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute delete full hash statement: database is locked
> > Nov 23 13:52:36 ossman.lkpg.cendio.se epiphany[3509]: Failed to execute clear hash prefix statement: attempt to write a readonly database
> > Nov 23 13:52:48 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute update threat statement: attempt to write a readonly database
> > Nov 23 13:52:59 ossman.lkpg.cendio.se epiphany[3509]: Failed to commit transaction on GSB database: disk I/O error
> > Nov 23 13:53:12 ossman.lkpg.cendio.se epiphany[3509]: Failed to execute update threat statement: attempt to write a readonly database
> > Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: attempt to write a readonly database
> > Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory
> > Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory
> > Nov 23 14:22:55 ossman.lkpg.cendio.se epiphany[3690]: Failed to execute insert hash prefix statement: out of memory
> 
> Those are the first lines from epiphany since upgrading to Fedora 32.

"disk I/O error" is interesting. Maybe time for a SMART test or a new disk? That's the sort of thing you never want to see. Take good backups?

(In reply to Michael Catanzaro from comment #19)
> 
> Hm, I don't think it's important enough for a F32 update, sorry. I'll
> backport it to 3.36 since you requested it, so it will be included if there
> are any further updates. For now, you can avoid this crash by going into
> privacy settings and disabling Block Dangerous Websites.
> 

Ah, thanks. I've disabled that now and hopefully it's the end of the crashes. :)

> 
> "disk I/O error" is interesting. Maybe time for a SMART test or a new disk?
> That's the sort of thing you never want to see. Take good backups?

My home directory is on NFS, so it should not be a disk problem. At least not a local one.

That said, there are no other log messages related to storage on either my machine or the NFS server. So again this message seems suspect.